How it works

Characters you cannot see

Unicode contains characters that render as nothing at all. A zero-width space (U+200B) occupies no width. The Unicode tag block (U+E0000–U+E007F) maps one to one onto printable ASCII, so a run of tag characters can carry an entire readable message that is completely invisible in any normal editor.

These are deterministic. There is no probability involved: a character is either U+200B or it is not. That makes them the one category this tool can find and remove with certainty.

Removal is context-aware, because several of these characters have legitimate jobs. A zero-width joiner is what holds an emoji sequence together. A zero-width non-joiner changes spelling in Persian and several Indic scripts. Tag characters build the Scotland, Wales and England flag emoji. Those are detected and kept, and the interface says which ones were kept and why.

Statistical watermarks in word choice

A model-level watermark does not add anything to the text. As a model generates, a secret keyed rule scores the candidate next tokens using the preceding context and nudges some of them very slightly upward. Each individual choice stays natural and invisible to a reader, but across a long enough passage the pattern of choices carries a statistical signal that a detector holding the matching key can test for.

Because the signal lives in the words themselves, it survives copying and pasting, and it can survive light editing. Heavy rewriting, paraphrasing, translation, or mixing with other writing may weaken it substantially.

What happens when detection is published

This is the thing most people arrive here wanting, so it is worth being exact about where it stands and what the plan is.

Today

No detector exists, not here and not anywhere else, because the method is not public. This tool reports Claude’s model watermark as detection unavailable rather than guessing, and the surface characters it does remove are named as characters, never as a watermark.

Detection

Anthropic says detection details will follow in forthcoming technical documentation. When it publishes a specification, an official detector, or an API that an implementation can be validated against, only one file changes here: the Anthropic adapter. The interface it plugs into is already built and tested.

Detection ships before anything else, because you cannot honestly claim to have removed something you were never able to confirm was there.

Removal

Once a mark is detectable, removal becomes a measurable question rather than a marketing one: run the detector, transform the text, run the detector again, and report what actually changed.

That is the work we intend to do, and the result is whatever the measurement says. A statistical mark lives in the word choices themselves, so the honest range of outcomes runs from reliably removed, through only weakened, to not removable without rewriting the text into something you did not write. Which of those it is cannot be known until there is a detector to measure against.

Until all of that has happened, every claim on this site about model-level watermarks stays exactly where it is: not detectable, therefore not removable. No reverse-engineered keys, no invented confidence scores, and no “watermark removed” message that a detector run cannot back up.

Metadata and signed provenance in files

Image and document files carry blocks of data alongside the picture. EXIF holds camera settings, GPS coordinates and timestamps. XMP and IPTC hold authorship and captions. ICC holds the colour profile. Each format stores them in a different place: JPEG uses APP marker segments, PNG uses text chunks, WebP uses RIFF chunks, SVG uses XML elements.

Removing them is container surgery. The file is rebuilt without those blocks while the compressed image data is copied across untouched, so the pixels are byte-identical. Two things are deliberately kept unless you ask otherwise: the EXIF orientation flag, because dropping it visibly rotates photos that viewers were auto-rotating, and the ICC profile, because dropping it shifts colour on wide-gamut images.

Content Credentials

C2PA provenance is a different thing again: a manifest containing signed claims about how a file was made, embedded in the file and cryptographically bound to its bytes. When a file carries one, this tool reads who signed it, what tool made it, whether it declares itself AI-generated, and whether the bytes still match the signature.

Two results that look similar mean very different things, so the interface keeps them apart. “The signature is valid and the bytes are unmodified” is a mathematical fact about the file. “The signer is on a recognised trust list” is a separate question, and many genuine files fail it simply because the certificate programme is young. A browser also cannot check certificate revocation, so that is reported as unchecked rather than quietly treated as fine.

Editing a signed file breaks its signature permanently. This tool will tell you before that happens, and it will never re-sign a file or imply that a modified one still carries valid provenance.

Common questions

Can this remove Claude’s text watermark?

Not yet, and neither can anything else right now. Anthropic has confirmed that Claude marks generated text at the model level, but has not published a detection method or a specification. Nothing here can detect that mark, so nothing here can honestly claim to remove it. What this tool does remove is invisible characters, which are a different thing entirely.

Are you going to build model-level watermark removal?

That is the plan, and it is blocked on the same thing everyone else is blocked on. Anthropic has said detection details are coming in forthcoming technical documentation. When that is published, detection is built here first, and removal is developed and measured against it: run the detector, transform the text, run the detector again, and publish what actually changed. Whether the result is reliable removal, partial weakening, or something that cannot be removed without rewriting the text is a question no one can answer until there is a detector to measure against.

Do invisible characters mean text was written by AI?

No. Zero-width characters and unusual spaces come from copying between editors, from web pages, from PDF exports, and from ordinary software. They are worth removing before you publish, but their presence says nothing about who or what wrote the text.

Does removing metadata reduce image quality?

No. Metadata is removed by rewriting the file container and copying the compressed image data across byte for byte. The picture is never decoded and re-encoded, so the result is pixel-identical to the original.

What happens to Content Credentials if I strip metadata?

The signature breaks permanently and cannot be restored. C2PA signs a hash of the file, so any change outside the manifest’s own region makes validation fail. This tool warns you before it happens and will never re-sign a file.

Are my files uploaded anywhere?

No. Every parser runs in your browser and there is no server-side processing, so there is nowhere for a file to be sent. This page and the focused tool pages do carry Google advertising, which contacts Google but never touches your file. The workspace under /app carries a second content security policy that forbids any outbound request at all, so there the guarantee is enforced by your browser rather than promised by us. The privacy page lists exactly which pages carry advertising.